![Sample our Humanities journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5939/TOC-Subject-Sample-2021-Humanities.jpg"})
ABSTRACT
Causes must be distinct from their effects. If the temperature in a room is 15°F, this can cause water pipes to freeze. However, the temperature’s being 15°F is not a cause of the temperature’s being below the freezing point. In general, conceptual, logical, mathematical, and other non-causal dependence relations should not be misclassified as causal. In this paper, I discuss how interventionist theories of causation can meet the challenge of distinguishing between (direct or indirect) causal relations and dependence relations that are due to non-causal factors. I survey several approaches to solving this problem, and propose a novel criterion based on the notion of supervenience. According to this criterion, variables are sufficiently distinct to qualify as the relata of a causal relation iff they do not have overlapping minimal supervenience bases.
Disclosure Statement
No potential conflict of interest was reported by the author.
Notes
1 There are differences between Woodward’s interventionist account and other causal modelling approaches (see, e.g., Spirtes, Glymour, and Scheines [Citation2000], and Hitchcock [Citation2001, Citation2007]). For reasons of space and ease of exposition, I will focus on Woodward’s account. However, my argument can be generalized to other versions of the causal modelling approach.
2 Strictly speaking, causal relations do not hold between variables; they hold between the events or properties represented by the values of the variables. For the sake of simplicity, I will often talk as if a variable X could be causally relevant to variable Y.
3 There has been a debate on whether variables not satisfying Independent Fixability can occur in the same causal model. Some authors argue that models containing variables that violate Independent Fixability are problematic (e.g. Eronen [Citation2012: 228], Yang [Citation2013], and Hoffmann-Kolss [Citation2014: 56]; see also Shapiro and Sober [Citation2007: 239–40]), whereas Woodward [Citation2015] argues that, under certain conditions, interventionism can handle models containing variables that violate Independent Fixability (for discussion, see Baumgartner [Citation2018]). In the present paper, I aim to stay neutral regarding this debate.
4 The example could be further generalized by replacing S6 with a variable Sk (where k is some even positive integer), taking the value 1 if a number <k is drawn and the value 0 otherwise.
5 A possible worry about this example is that variables such as S and M raise further problems, since it seems that what hypothetical interventions would have to be carried out in order to investigate their causal influence on other variables depends on the context (Woodward [Citation2003: 114–17]). However, for the purposes of the present argument, it will be sufficient to presuppose that there are causal models containing variables such as S and M. Since that presupposition seems to be generally accepted among proponents of interventionist causation, I will set aside this problem.
6 This is, of course, an adapted version of the classical notion of strong supervenience: see, e.g., Kim [Citation1984: 165].
7 This holds at least if the values of X and Y are always defined. Then, necessarily, whenever X assumes some value xi, there is a value of Y, yj, such that Z = (xi & yj), and, necessarily, if Z = (xi & yj) then X = xi. (And an analogous claim holds for Y.)
8 This does not cover mixed cases in which the values of X represent properties and the values of Z represent events or vice versa. However, since it is usually assumed that only entities belonging to the same ontological category can stand in causal relations to each other, I will leave aside this complication. Depending on the underlying metaphysics, the values of variables could also represent facts, propositions, or states of affairs. However, since these entities are also constituted by properties, it would not require any substantial modifications of the proposed criterion to cover them as well.
9 Note that condition (ii) does not contradict the observation that the values of variables in causal models can stand for disjunctive properties or other non-natural properties (see section 4). The requirement that the values of a variable must be comparatively natural concerns only the variables contained in minimal supervenience bases, and is perfectly compatible with the assumption that the values of causal variables stand for non-natural properties (or non-natural events).
10 Note that most variables will have several minimal supervenience bases. Consider, for instance, the following variable:
T*: r iff the temperature in room A is r° F (where r is some real number)
11 A previous version of this paper was presented at the conference of the Society for the Metaphysics of Science in 2017. I would like to thank Dmitri Gallow, my commentator at this conference, for his very constructive feedback that helped me to bring the argument to its present form. In addition, I have benefitted from discussions with various people and audiences—in particular, Jens Harbecke, Michael Hicks, Andreas Hüttemann, Siegfried Jaag, Lena Kästner, Christian Löw, Matthias Rolffs, Jonas Werner, and the members of the eidos Centre for Metaphysics in Geneva. I would also like to thank two anonymous referees from this journal for providing extremely detailed and helpful comments on earlier versions of the manuscript.